Makes AI coding agents think critically, reason deeply, and act with integrity. No hallucination. No excuses. No sycophancy.
npx skills add vtino17/deep-mind-skillFor Claude Code:
/plugin marketplace add vtino17/deep-mind-skillOnce installed, just include any of these in your prompt:
Commands:
/think-deeper <question>— Full 5-stage reasoning/critical <claim>— Adversarial review/verify <statement>— Evidence gate/deep-dive <problem>— Root cause analysis/first-principles <concept>— Break down from fundamentals
Natural language:
- "think deeper: why does this keep failing?"
- "critical analysis: is this architecture correct?"
- "verify: does this library support Python 3.13?"
User: /think-deeper why is port 443 open on 10.0.0.1?
Agent:
[DEEP-MIND] Applying 5-stage reasoning...
1. VERIFY — Port 443 open means HTTPS service running
2. GROUND — Nmap scan result: 443/tcp open https
3. REASON — HTTPS requires TLS certificate + web server
4. CHALLENGE — Could be non-HTTP service masquerading on 443
5. DECIDE — Confirmed: TLS handshake successful, subject: *.example.com
- 5-Stage Reasoning: VERIFY, GROUND, REASON, CHALLENGE, DECIDE
- Anti-Hallucination: Blocks fabricated code and imaginary APIs
- Evidence Gate: Every claim must cite real sources
- Anti-Sycophancy: Rejects user flattery, tells the truth
- Zero Dependencies: Pure markdown, no pip/npm install needed
- Agent-Agnostic: Works with Claude Code, Codex, Cursor, Windsurf, Aider, Cline, Continue, Roo Code, Gemini CLI, Copilot CLI, OpenCode
- Plugin support: OpenCode slash commands
Every time your agent receives a task, deep-mind intercepts and runs it through five stages:
1. VERIFY Check the request against known facts
2. GROUND Anchor every claim to concrete sources
3. REASON Step-by-step logical deduction
4. CHALLENGE Stress-test with counter-evidence
5. DECIDE Output only what survives all gates
If any stage fails, the agent reports the gap instead of guessing.
| Agent | Status | Commands |
|---|---|---|
| Claude Code | Full | /think-deeper, /critical, /verify |
| Codex CLI | Full | @deep-mind |
| Cursor | Full | /think-deeper |
| OpenCode | Full | /think-deeper, /critical, /verify |
| Windsurf | Full | Manual skill reference |
| Gemini CLI | Full | Manual skill reference |
| Copilot CLI | Supported | AGENTS.md reference |
| Cline, Continue, Aider, Roo Code | Supported | Manual skill reference |
| Repo | Stars | Description |
|---|---|---|
| kage | AI-powered security scanner | |
| tools | 85+ penetration testing tools | |
| taskcapsule | Task context manager | |
| network-security-lab | Enterprise network security lab | |
| mikrotik-hardening | RouterOS security configs | |
| incident-response-playbooks | IR playbooks and scripts | |
| vuln-scanner | Vulnerability scanner | |
| pcap-forensics | PCAP analysis tool | |
| network-automation-toolkit | Multi-vendor automation |
MIT — use, modify, share freely.